threads / patch / 62792

patcht8002-blame: simplify padding generation in blank boundary tests

Subject: [PATCH] t8002-blame: simplify padding generation in blank boundary tests

## tl;dr

6 messages between Jan 11, 2025 and Jan 21, 2025. Diffs are folded; open one to read it.

replies: 5people: 3as markdown or json

Jan Palus· Jan 11, 2025, 23:11 UTC · lore

Fixes compatibility with mksh as well: $ mksh -c 'printf "%0.s" ""' printf: %0.s: invalid conversion specification

Fixes: e7fb2ca945 ("builtin/blame: fix out-of-bounds write with blank boundary commits")
Signed-off-by: Jan Palus <jpalus@fastmail.com>
---
 t/t8002-blame.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
Show changes to t/t8002-blame.sh +2 −2
diff --git a/t/t8002-blame.sh b/t/t8002-blame.sh
index 1ad039e123..e98993276a 100755
--- a/t/t8002-blame.sh
+++ b/t/t8002-blame.sh
@@ -138,7 +138,7 @@ test_expect_success 'blame --abbrev -b truncates the blank boundary' '
 	# Note that `--abbrev=` always gets incremented by 1, which is why we
 	# expect 11 leading spaces and not 10.
 	cat >expect <<-EOF &&
-	$(printf "%0.s " $(test_seq 11)) (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
+	$(printf "%11s" "") (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
 	EOF
 	git blame -b --abbrev=10 ^HEAD -- abbrev.t >actual &&
 	test_cmp expect actual
@@ -146,7 +146,7 @@ test_expect_success 'blame --abbrev -b truncates the blank boundary' '
 
 test_expect_success 'blame with excessive --abbrev and -b culls to hash length' '
 	cat >expect <<-EOF &&
-	$(printf "%0.s " $(test_seq $hexsz)) (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
+	$(printf "%${hexsz}s" "") (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
 	EOF
 	git blame -b --abbrev=9000 ^HEAD -- abbrev.t >actual &&
 	test_cmp expect actual

-- 
2.48.0
Patrick Steinhardt· Jan 13, 2025, 12:36 UTC · re: Jan Palus · lore

Re: [PATCH] t8002-blame: simplify padding generation in blank boundary tests

On Sun, Jan 12, 2025 at 12:11:07AM +0100, Jan Palus wrote:
Show 5 quoted lines
> Fixes compatibility with mksh as well:
> $ mksh -c 'printf "%0.s" ""'
> printf: %0.s: invalid conversion specification
> 
> Fixes: e7fb2ca945 ("builtin/blame: fix out-of-bounds write with blank boundary commits")

We don't typically use Fixes tags in our project, but instead embed the commit into the commit message with `git log --format=reference -1` together with a description.

The subject can also be adjusted a bit: we use to just write the test number, and the important aspect is not that we simplify the padding generation, but that we make it more portable.

So, my suggestion would be:
    t8002: fix unportable printf formatting directives
    In e7fb2ca945 (builtin/blame: fix out-of-bounds write with blank
    boundary commits, 2025-01-10), we have introduced two new tests that
    expect a certain amount of padding. This padding is generated via
    printf using the "%0.s" formatting directive. That directive is
    non-portable and not understood by for example mksh, breaking these
    tests on platforms using that shell.
    Fix this issue by using "%${N}s" instead, which is already being
    used in t5300 and thus portable enough for us.
Show 18 quoted lines
> Signed-off-by: Jan Palus <jpalus@fastmail.com>
> ---
>  t/t8002-blame.sh | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/t/t8002-blame.sh b/t/t8002-blame.sh
> index 1ad039e123..e98993276a 100755
> --- a/t/t8002-blame.sh
> +++ b/t/t8002-blame.sh
> @@ -138,7 +138,7 @@ test_expect_success 'blame --abbrev -b truncates the blank boundary' '
>  	# Note that `--abbrev=` always gets incremented by 1, which is why we
>  	# expect 11 leading spaces and not 10.
>  	cat >expect <<-EOF &&
> -	$(printf "%0.s " $(test_seq 11)) (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
> +	$(printf "%11s" "") (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
>  	EOF
>  	git blame -b --abbrev=10 ^HEAD -- abbrev.t >actual &&
>  	test_cmp expect actual

Okay, makes sense. And as mentioned, we already have such a use of printf in t5300, so it should be portable enough for our use case.

Thanks!
Patrick
Junio C Hamano· Jan 13, 2025, 15:15 UTC · re: Patrick Steinhardt · lore

Re: [PATCH] t8002-blame: simplify padding generation in blank boundary tests

Patrick Steinhardt <ps@pks.im> writes:
Show 28 quoted lines
> On Sun, Jan 12, 2025 at 12:11:07AM +0100, Jan Palus wrote:
>> Fixes compatibility with mksh as well:
>> $ mksh -c 'printf "%0.s" ""'
>> printf: %0.s: invalid conversion specification
>> 
>> Fixes: e7fb2ca945 ("builtin/blame: fix out-of-bounds write with blank boundary commits")
>
> We don't typically use Fixes tags in our project, but instead embed the
> commit into the commit message with `git log --format=reference -1`
> together with a description.
>
> The subject can also be adjusted a bit: we use to just write the test
> number, and the important aspect is not that we simplify the padding
> generation, but that we make it more portable.
>
> So, my suggestion would be:
>
>     t8002: fix unportable printf formatting directives
>
>     In e7fb2ca945 (builtin/blame: fix out-of-bounds write with blank
>     boundary commits, 2025-01-10), we have introduced two new tests that
>     expect a certain amount of padding. This padding is generated via
>     printf using the "%0.s" formatting directive. That directive is
>     non-portable and not understood by for example mksh, breaking these
>     tests on platforms using that shell.
>
>     Fix this issue by using "%${N}s" instead, which is already being
>     used in t5300 and thus portable enough for us.

Is "%.0s" really not portable, or is it just mksh being a bit lacking?

"That directive non-portable ..." -> "Some implementations (e.g. one that is built into mksh) does not support the precision to be 0 (i.e. "%.0" before the "s" conversion)"

Other than that, your version is easy to read and understand.
Show 8 quoted lines
>> -	$(printf "%0.s " $(test_seq 11)) (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
>> +	$(printf "%11s" "") (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
>>  	EOF
>>  	git blame -b --abbrev=10 ^HEAD -- abbrev.t >actual &&
>>  	test_cmp expect actual
>
> Okay, makes sense. And as mentioned, we already have such a use of
> printf in t5300, so it should be portable enough for our use case.
Thanks for reviewing, and thanks, Jan, for noticing and fixing.
Junio C Hamano· Jan 18, 2025, 01:09 UTC · re: Junio C Hamano · lore

Re: [PATCH] t8002-blame: simplify padding generation in blank boundary tests

Junio C Hamano <gitster@pobox.com> writes:
Show 33 quoted lines
>> So, my suggestion would be:
>>
>>     t8002: fix unportable printf formatting directives
>>
>>     In e7fb2ca945 (builtin/blame: fix out-of-bounds write with blank
>>     boundary commits, 2025-01-10), we have introduced two new tests that
>>     expect a certain amount of padding. This padding is generated via
>>     printf using the "%0.s" formatting directive. That directive is
>>     non-portable and not understood by for example mksh, breaking these
>>     tests on platforms using that shell.
>>
>>     Fix this issue by using "%${N}s" instead, which is already being
>>     used in t5300 and thus portable enough for us.
>
> Is "%.0s" really not portable, or is it just mksh
> being a bit lacking?
>
> "That directive non-portable ..." -> "Some implementations (e.g.
> one that is built into mksh) does not support the precision to be 0
> (i.e. "%.0" before the "s" conversion)"
>
> Other than that, your version is easy to read and understand.
>
>>> -	$(printf "%0.s " $(test_seq 11)) (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
>>> +	$(printf "%11s" "") (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
>>>  	EOF
>>>  	git blame -b --abbrev=10 ^HEAD -- abbrev.t >actual &&
>>>  	test_cmp expect actual
>>
>> Okay, makes sense. And as mentioned, we already have such a use of
>> printf in t5300, so it should be portable enough for our use case.
>
> Thanks for reviewing, and thanks, Jan, for noticing and fixing.

Sorry, as Jan is not a list regular, perhaps I should have communicated more carefully when I said "Thanks".

The above message from me with "Thanks" does not mean that the patch is now settled. There are suggested improvements pending that needs to be incorporated before the patch becomes acceptable to our tree.

Anybody can help that "further polishing as suggested" step, and when the patch is left in limbo for too long, I might step in to do it myself (when I have no other better things to do), but it is customary around here that the original patch submitter does so.

Thanks.
Jan Palus· Jan 20, 2025, 10:47 UTC · re: Junio C Hamano · lore

Re: [PATCH] t8002-blame: simplify padding generation in blank boundary tests

On 17.01.2025 17:09, Junio C Hamano wrote:
Show 18 quoted lines
> Junio C Hamano <gitster@pobox.com> writes:
> 
> >> So, my suggestion would be:
> >>
> >>     t8002: fix unportable printf formatting directives
> >>
> >>     In e7fb2ca945 (builtin/blame: fix out-of-bounds write with blank
> >>     boundary commits, 2025-01-10), we have introduced two new tests that
> >>     expect a certain amount of padding. This padding is generated via
> >>     printf using the "%0.s" formatting directive. That directive is
> >>     non-portable and not understood by for example mksh, breaking these
> >>     tests on platforms using that shell.
> >>
> >>     Fix this issue by using "%${N}s" instead, which is already being
> >>     used in t5300 and thus portable enough for us.
> >
> > Is "%.0s" really not portable, or is it just mksh
> > being a bit lacking?
Contrary to other shells mksh does not have printf builtin:

$ mksh -c 'type printf' printf is a tracked alias for /bin/printf

so it uses printf from coreutils. This version however interprets "0" as a flag marking "s"/"c" conversion specifiers as not allowed:

https://git.savannah.gnu.org/gitweb/?p=coreutils.git;a=blob;f=src/printf.c;h=2a73bb7fed892347eafb40f497ce5080f511fc9b;hb=v9.6#l586
Show 5 quoted lines
> > "That directive non-portable ..." -> "Some implementations (e.g.
> > one that is built into mksh) does not support the precision to be 0
> > (i.e. "%.0" before the "s" conversion)"
> >
> > Other than that, your version is easy to read and understand.

Note that original version was "%0.s" in which there is some ambiguity whether "0" is a flag or field width and not "%.0s" in which "0" indeed would mean precision.

Show 22 quoted lines
> >>> -	$(printf "%0.s " $(test_seq 11)) (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
> >>> +	$(printf "%11s" "") (<author@example.com> 2005-04-07 15:45:13 -0700 1) abbrev
> >>>  	EOF
> >>>  	git blame -b --abbrev=10 ^HEAD -- abbrev.t >actual &&
> >>>  	test_cmp expect actual
> >>
> >> Okay, makes sense. And as mentioned, we already have such a use of
> >> printf in t5300, so it should be portable enough for our use case.
> >
> > Thanks for reviewing, and thanks, Jan, for noticing and fixing.
> 
> Sorry, as Jan is not a list regular, perhaps I should have
> communicated more carefully when I said "Thanks".
> 
> The above message from me with "Thanks" does not mean that the patch
> is now settled.  There are suggested improvements pending that needs
> to be incorporated before the patch becomes acceptable to our tree.
> 
> Anybody can help that "further polishing as suggested" step, and
> when the patch is left in limbo for too long, I might step in to do
> it myself (when I have no other better things to do), but it is
> customary around here that the original patch submitter does so.

I was about to follow-up but didn't find time. Sorry it took so long. I will post v2 shortly.

Junio C Hamano· Jan 21, 2025, 20:10 UTC · re: Jan Palus · lore

Re: [PATCH] t8002-blame: simplify padding generation in blank boundary tests

Jan Palus <jpalus@fastmail.com> writes:
> Note that original version was "%0.s" in which there is some ambiguity
> whether "0" is a flag or field width and not "%.0s" in which "0" indeed
> would mean precision.

Ah, I missed that part. Also thanks for filling in the "printf from coreutils is the one that has issues with the code".

Show 7 quoted lines
>> Anybody can help that "further polishing as suggested" step, and
>> when the patch is left in limbo for too long, I might step in to do
>> it myself (when I have no other better things to do), but it is
>> customary around here that the original patch submitter does so.
>
> I was about to follow-up but didn't find time. Sorry it took so long.
> I will post v2 shortly.
Thanks.

← back to recent threads